What Is an Authenticity Score in Influencer Vetting?
An authenticity score is an internal rating an operator assigns to a creator account to express how much of its audience and engagement appear to be real people, as opposed to purchased followers, automated accounts, or low-quality engagement-farm activity. It is a category of vetting check rather than a standardized industry metric: no platform or standards body publishes an official scale, so every authenticity score inherits the rubric of the team that computed it and should be read with that rubric attached.
Turn this page into a campaign built for your brand.
Connect Superdirector to ChatGPT and it adds current social-video references, public-video analysis, Brand DNA, and a ready-to-send campaign brief.
By Bell Chen, founder.
The reason the check exists is arithmetic. An influencer quote is a price per access to an audience, and the audience is the one input the brand cannot verify from the media kit. The market context makes the stakes concrete: per the Collabstr 2026 report (collabstr.com), the average paid collaboration ran $193 on Instagram and $255 on YouTube across 21,000+ priced deals, with engagement averaging 2% on TikTok and 6% on YouTube, so an account whose real engagement is a fraction of its reported engagement is being bought at a multiple of its true rate. An authenticity score is the operator answer: turn the signals of a real audience into a weighted score, run it on every candidate before the brief, and let the sub-scores, not the follower count, carry the decision.
What an authenticity score actually measures
It measures evidence quality, in two layers. The follower layer asks whether the audience appears to exist: growth shape against posting cadence, follower-to-engagement ratio stability, and audience geography against the markets the brand sells in. The engagement layer asks whether the attention is real: comment specificity, the presence of saves and shares rather than likes alone, and whether engagement concentrates on posts the way organic attention does.
What it does not measure is the creator. A score is a snapshot of evidence on an account at a point in time, assembled from public signals and whatever audience report the creator supplies. Teams that treat the number as a character judgment end up in the two failure modes the check is meant to prevent: paying for audiences that do not exist, and accusing real creators on thin evidence. The score works as a reason to ask for more evidence, not as a verdict.
How a rubric version is built
The house version used across this site scores four measured signals on a 100-point rubric: US-audience percentage, engagement rate against format norms, fake-follower indicators, and average recent Reel pulls, with the weights tuned per campaign. In one house operating record from February 2026, a creator brief drafted before the rubric existed was re-scored at 29 out of 100 and would have cost $571 per thousand impressions; the same account class, vetted, is the reason the rubric was adopted. The full weights, thresholds, and a worked example are written out in the how-to guide on vetting with the 100-point rubric.
No tool output is a score by itself. Third-party audience estimates, platform analytics, and marketplace data all feed the rubric, and the operator assigns the sub-scores after reading the evidence. Any tool that claims to hand down a single authenticity number without showing its weights is a black box, and a black box score cannot be defended to a client or improved after a bad outcome.
What the reference numbers say
The benchmark that matters for the engagement layer is the format spread, because a raw engagement rate means nothing without its format and size context. Per the Collabstr 2026 report (collabstr.com), average engagement runs 6% on YouTube, 5% on Instagram Reels, 4.5% on YouTube Shorts, and 2% on TikTok. An account running far above its format norm with thin comment quality is a stronger fraud candidate than one running slightly below it, because purchased engagement tends to concentrate in likes.
The follower layer has no published benchmark worth citing, which is itself the finding: no platform publishes fake-follower rates, so any percentage a vendor quotes for an account is a model estimate. The honest practice is to score the named indicators, show the evidence for each, and hold the conclusion at the confidence the evidence supports, which is also how the house rubric phrases its output.
How to run the check on a candidate
Pull four facts before reading anything qualitative: audience geography, engagement rate by post for the last ten posts, follower count over time, and average recent video pulls. Every one of them is checkable from public pages or from an audience report the creator provides, and a creator who declines to provide one that is standard for their tier has answered a question already.
Then read the comment fields by hand on the last five posts. Template phrases, sequences of bare emoji, and commenters with no visible activity of their own are the pattern purchased engagement produces. Genuine audiences argue, ask about specifics, and mention details the post did not feature, because they own the product context the creator is describing.
Score the signals on a written rubric and keep the sub-scores. The number is only as useful as its ability to explain a decision later, which is also why the sub-scores, not the total, are what a client review actually reads.
Common mistakes
The most common mistake is single-signal verdicts. A follower spike can be one genuine post taking off, a low engagement rate can be a dormant-but-real audience, and a template-heavy comment field can be a niche with a young audience. The rubric exists so that several weak signals, not one, move the decision.
The second mistake is trusting tool numbers without weights. Any authenticity percentage produced by a third-party tool is a model output, and the model is not disclosed. The defensible practice is scoring named signals you can explain, and treating tool outputs as one input among several.
The third mistake is running the check after the brief instead of before it. Once a brief is written, the sunk cost of the drafting work biases the read of the evidence, which is how 29-out-of-100 accounts end up in negotiations. The rubric runs first, and the brief is written for the accounts that clear it.
Where a planning-first tool fits
Inside Superdirector, the vetting work lives as a written scorecard on the planning canvas: the four measured signals, the sub-scores, and the evidence notes travel with the brief so the next campaign inherits the judgment rather than redoing it. The scoring is done by the operator, not automated by the tool; what the planning layer contributes is keeping the rubric, the evidence, and the brief attached to each other.
Disclosure by Bell Chen, founder of Superdirector: the planning and vetting-note features mentioned in this piece are part of the product I build, and they do not auto-score creators; the rubric is applied by a person. The February 2026 score cited is a house operating record, and the market figures are sourced from the linked marketplace report.
Frequently asked questions
What signals suggest an account has fake followers?
Is there a standard scale for authenticity scores?
What does a low authenticity score mean in practice?
How do you check engagement quality without paid tools?
Get a campaign brief built for your brand in 30 seconds
Vet creators with a written rubric before the brief goes out